EP3577626B1 - Method for dimensional x-ray measurement, in particular by computed tomography, and x-ray computed tomography scanner - Google Patents

Description

The invention relates to a method for dimensional measurement using computed tomography according to the preamble of claim 1. According to a second aspect, the invention relates to an x-ray computer tomograph according to the preamble of the independent claim.

Radiography describes the process in which a test object is irradiated with X-rays and the intensity of the radiation in the radiation path behind the test object is measured as an image. Computed tomography is an extension of determining a three-dimensional image by rotating and, if necessary, moving the test object relative to the detector and the X-ray source. The three-dimensional volume image is calculated from the images in the various angular positions, the projections, using complex mathematical processes. In industrial computed tomography, microfocus X-ray sources and flat panel detectors are preferably used. This means that the relative measurement uncertainty when determining dimensions is currently limited to 1 × 10-4, as international comparative measurements show. It is desirable to achieve a smaller measurement uncertainty.

Disadvantages in industrial computed tomography arise from the fact that the X-ray sources used emit broadband bremsstrahlung and characteristic radiation, ie non-monochromatic X-ray radiation. In addition, the maximum power for a given source spot diameter is limited. Although it would be possible to use, for example, a synchrotron radiation source or a free-electron laser, which do not have these disadvantages, this requires one dedicated large-scale research facility that cannot be used for industrial measurement tasks in routine operation.

The invention relates to the use of non-monochromatic X-ray radiation. Advantageously, with the aid of the invention, comparable good results are made possible as when using monochromatic X-ray radiation. As a practical measure, strong pre-filtering of the X-rays has already been used to make the spectrum narrower at the expense of a lower X-ray intensity. This contradicts the fact that industrial computed tomography also generally requires short measurement times in order to be economical. The method according to the invention can enable less pre-filtering, thus reducing the measuring time and is therefore more economical. The invention is also based on the object of reducing the measurement uncertainty in computed tomography, in particular industrial computed tomography, which is systematically adversely affected by the broadband nature of the x-ray source.

From the article Influence of source movement on reproducibility and probing deviations in X-ray computed tomography by Weiss et al.: Proceedings Industrial Computed Tomography Conference, Wels, Austria, 2010 , it is known that the thermal expansion of the target due to the bombardment with the electrons can lead to a displacement of the focal spot. The article states that changing the operating parameters such as acceleration voltage and beam current due to heating and the associated movement of the source location can lead to errors, whereas beam hardening, artifacts of the reconstruction via unknown effects play no role in the case under investigation. Even with negligibly small fluctuations in the focal spot, systematic measurement uncertainties occur with this method.

From the article Depth-of-interaction estimates in pixelated scintillator sensors using Monte Carlo techniques by Sharma et al. in Nuclear Instruments and Methods in Physics Research A 841 (2017), pages 117 - 123 describes how the optical paths of photons in the scintillation detector depend on parameters such as surface roughness and absorption at the surface and the solid.

It is assumed that the entire energy of a photon is emitted at the first scattering event; Compton scattering is not modeled either. The effect that different, non-rotationally symmetrical intensities result on neighboring pixels when the beam enters obliquely, depending on the spectral hardening of the beam, is not described.

In the US 8,208,603 B2 describes how the anode angle affects the spectrum and the intensity distribution of a useful X-ray radiation cone of an X-ray tube and how this effect can be corrected. The effect of beam hardening on changing the effective penetration depth or the effective source location is not described.

The object of the invention is to reduce the measurement uncertainty in computed tomography (CT), in particular in industrial computed tomography.

The invention solves the problem by a method having the features of claim 1.

According to a second aspect, the invention solves the problem with an X-ray computer tomograph having the features of the independent claim.

The invention is based on the finding that the gray values of neighboring pixels of the detector due to absorption by the test object in radiographic recordings, in particular in cone beam geometry, with a broad spectrum, in contrast to almost parallel and/or monochromatic X-ray radiation, due to radiation-physical characteristics of the source and the detector show directional crosstalk. The resulting local change in size scaling in the radiographic image is transferred to a reconstructed CT image, i.e. a three-dimensional density image of the specimen that is calculated from the pixel-dependent intensity data for different orientations of the specimen relative to the X-ray source and detector. The test object surfaces are calculated from the density image and dimensions are determined, which are ultimately changed by the directed crosstalk in the gray value image. The local radiographic magnification factor gives is the quotient of the distance from the effective source point to the effective detection location as the numerator to the distance between the effective source point and the test object as the denominator. In CT, the location of the specimen is given by the position of the axis of rotation, the effective source point and detection location are given by the expected value of the absorption or emission averaged over the spectrum absorbed in the specimen on the way from the source point to the detection location.

It is known that broadband X-rays show a beam hardening when passing through the test object. This is to be understood as meaning the phenomenon that X-rays of higher energy and therefore shorter wavelengths are normally absorbed to a lesser extent than X-rays of lower energy and therefore longer wavelengths. The X-ray spectrum thus changes with a relative increase in the proportion of "hard", i.e. higher-energy photons.

The inventors have recognized that the effective penetration depth of the X-ray radiation in the detector is greater, the stronger it is absorbed in the test object. This effect is negligible as long as the X-ray beam hits the detector perpendicularly. If the detector is curved in such a way that the center of the circle of curvature coincides with the source point of the X-ray source, this effect does not result in a measurement error.

However, curved planar detectors are difficult to manufacture and are therefore generally not used. However, this means that at those points where the X-ray beam hits the detector at an angle deviating from 90°, the X-ray photons are detected on average the further out, the harder the X-ray radiation is.

Advantageously, the invention can be combined with the already known beam hardening correction, in which pixel by pixel their gray values are corrected with a non-linear function in order to take into account the deviation of the exponential attenuation of the intensity with the absorber thickness, known as de Beer's law. By applying the correction of the invention in addition to the known beam hardening correction, the radiographic image more closely matches the image produced by an ideal, i.e. very thin, detector at a given source-to-detector spacing would have measured - a very thin detector does not have sufficient absorption for X-rays. This reduces the measurement uncertainty.

Another advantage is that this correction is mathematically simple and can therefore be carried out quickly without having to provide large computing capacity.

A further advantage is that the three-dimensional model can be calculated from the corrected pixel-dependent intensity data in the same way as with uncorrected pixel-dependent intensity data. In other words, existing software can be used.

It is also advantageous that the measurement time can be reduced since, as already explained above, there is no need for strong pre-filtering of the X-ray radiation. Because of the correction of the effective penetration depth on the detector and/or a shift in the effective source location, X-ray radiation can be used that has a greater spectral width. In other words, a weaker or no pre-filter can be used compared to previous measurement strategies. This shortens the measurement time.

Another advantage is that larger specimens can be measured. With a typical acceleration voltage of U=225 kilovolts, the dimensions of test specimens can be sensibly measured if the material thickness is a maximum of 75 millimeters if they are made of aluminum or a maximum of 15 millimeters if they are made of steel. These thicknesses correspond to 95% absorption when the radiation has been prefiltered with 2 mm copper. It has been empirically determined that the measurement uncertainty already increases with absorption of more than 50%, without the cause of this effect having been correctly identified. Since the influence of the spectrum dependency on source and detector position, which causes a local change in the shadow cast, is now known and can therefore be corrected, larger absorptions can be tolerated. In other words, specimens with larger dimensions can be dimensionally measured.

A relative distance measurement uncertainty of below 3×10 ^-5 can be achieved by the correction according to the invention. In particular for easily deformable objects, such as plastic products, this measurement uncertainty is in the measurement range that is also achieved by tactile measurements. There is therefore a justified expectation that computed tomography based on the invention can increasingly replace tactile measurements in the future.

In addition to the aspect described above that the beam hardening in the test object and pre-filter influences the position of the shadow image in the detector material, another aspect that can be understood geometrically from this contributes to the systematic measurement uncertainty. If an X-ray beam hits the detector that has been greatly weakened, i.e. hardened, its photons - as explained above - have a higher average energy. This in turn means that the photons are more likely to come from a source location on the target that is close to where the electron beam enters the surface, because the electrons still have a particularly high kinetic energy there.

Accordingly, photons that strike the detector at a location where a high intensity is measured have a relatively greater likelihood of coming from a location on the target where the electrons were of lower energy. For example, this can be a point deeper below the surface of the target. If the point of origin of the X-ray radiation is approached as a point, the position of this point and thus the distance between this point and the detector on the one hand and the distance between this point and the test object on the other hand depend on the intensity measured in the detector. Correcting this influence on the radiographic image and thus reducing the measurement uncertainty is an independent subject matter of the invention and is also preferably a preferred embodiment of the aspect of the invention described above. In other words, the shift in the effective source location is corrected.

In the context of the present description, the use of an indefinite article always means that either exactly one or more of the corresponding objects are present. Exactly one point source can be present, but also two or more point sources.

A point source is understood to mean, in particular, an x-ray source in which the source location of the x-ray radiation can be viewed as point-like to a sufficiently good approximation. This means in particular that this approximation leads to a systematic measurement uncertainty of at most 10 ^-5 . In particular, when the electron beam strikes a target of the X-ray source, it has a diameter (diameter at half the maximum value) of at most 0.5 millimeters, in particular at most 0.1 millimeters. Preferably, the x-ray radiation is in the form of a cone beam or a fan beam.

Computed tomography is also understood to mean, in particular, a laminography.

A dimension is also understood to mean, in particular, a geometric dimension or a geometric feature.

X-rays are non-monochromatic. It is preferably bremsstrahlung with a proportion of characteristic radiation of the target material. It is favorable if the x-ray radiation is produced by irradiating a target with electrons, with an energy of the electrons preferably being between 20 and 600 kiloelectron volts, preferably between 60 and 225 kiloelectron volts.

Moving the specimen relative to the X-ray source is understood to mean moving when the X-ray source is not moving, moving the X-ray source when the specimen is not moving, or moving both.

The feature that the pixel-dependent intensity data is corrected for the influence of a beam hardening-related change in the effective penetration depth means that the intensity data are changed in such a way that the effect is corrected that the intensity measured by a pixel correlates positively with the effective penetration depth, that is, a high intensity indicates a high effective penetration depth.

The feature that the shift in the effective source location on a target of the x-ray source is corrected means in particular that the shift in the effective source location caused by the beam hardening is corrected.

The detector is preferably a scintillation detector or a volume-absorbing semiconductor detector. It is favorable if this has microcolumns made of scintillating material. The scintillating material is, for example, cesium iodide doped with thallium or contains a gadolinium, tungsten and/or a lanthanum compound.

A thickness of a detector layer, in particular the height of the microcolumns, is preferably at least 400 micrometers, in particular at least 500 micrometers. In this case, the influence of the spectrum dependence of the effective penetration depth of the shadow image of the test object is particularly large.

However, the shift in the source location is usually not only due to beam hardening. A shift in the source location also results from the combination of the influence of the beam hardening and the spectrally variable X-ray emission via the path of the radiation-generating electron beam in the X-ray target. This influence is preferably also corrected

By the steps: for at least a majority of the pixels, in particular all pixels, (i) determining a zero-point distance of the pixel from an optical axis, (ii) determining the intensity of the x-ray radiation measured by the pixel, (iii) assigning a corrected position , which depends on the zero point distance and the intensity, and (iv) calculating corrected pixel-dependent intensity data from all corrected positions and the associated intensities, a three-dimensional model of the test object can be calculated from this corrected pixel-dependent intensity data. In particular, this calculation does not differ from a method that is known from the prior art.

It should be noted that determining the intensity of the x-ray radiation measured by the pixel also includes determining the intensity of the absorbed x-ray radiation measured by the pixel.

The optical axis is understood to mean in particular that straight line along which an imaginary x-ray beam runs from the x-ray source to the detector, this x-ray beam running perpendicular to a detector surface along which the detector extends.

The effective shift in source and detector location caused by the absorption spectrum has no influence on the detected position of an imaginary X-ray beam in relation to a detector pixel, provided the pixel in question lies on the optical axis. The greater the distance between a pixel and the optical axis, the greater the measurement location deviation caused by the shifts.

The two corrections, namely the source location and the penetration depth in the detector, can be different in different areas of the intensity data.

Determining the intensity means, in particular, that a measured value is recorded that describes the intensity of the x-ray radiation. For example, the intensity determined in this way is a grayscale value that encodes the intensity measured by the corresponding pixel.

The correction of the pixel-dependent intensity data is preferably carried out in such a way that the corrected position, the zero point and the original position of the corresponding pixel lie on a line. The zero point is the point on the detector at zero distance from the optical axis. In other words, if the detector is viewed in a polar coordinate system, the correction for the influence of the penetration depth changes the distance coordinate, but not the azimuthal angle.

To compensate for the effect of the effective source and/or detector location shift for each pixel, the effect of the negative sign effect on the image is performed such that a difference distance Δr between the zero point distance of the corrected position and the zero point distance the uncorrected position r is essentially calculated from a product of the distance r, the absorbed intensity (I ₀ -I) measured by the respective pixel and a Intensity correction parameter k and a constant c dependent on the filter and sample material: Δr = r * ((I ₀ -I) * k + c).

A radial displacement of the pixels proportional to the distance r describes a change in the geometric magnification factor of an industrial CT with point source and flat panel detector. The constant c describes a change in magnification for the entire image due to the position of the absorption image of a thin layer of the test sample material in the detector. This depends on the spectrum of the X-ray source, ie in particular the acceleration voltage, target material, filter material and thickness. This intensity correction parameter k depends on the material under test, the source conditions, in particular acceleration voltage and target material, and the detector, in particular material and thickness.

The intensity correction parameter, on the other hand, is not dependent on the zero-point distance r itself, at least in the linear order of the zero-point distance. In particular, the intensity correction parameter is not dependent on the intensity to a good approximation, as experimental results suggest. He describes empirically that when the absorbed intensity changes, the beam hardens and thus the depth of penetration into the detector material in the vicinity of the corresponding pixel changes. The same applies to the source location shift. This leads locally to a change in magnification at this point and thus to a shift in intensity from this pixel to neighboring ones. The parameters k and c can be assumed to be different for different areas of the image if different test specimen materials contribute to the absorption in these areas.

The formula above applies to the radially symmetrical case, which is caused by the detector location shift δ. This symmetry is broken when the source location is shifted, insofar as the source location moves along an inclined target with a component parallel to the detector. This can be taken into account by adding an additional shift in the Y direction (in the direction of the target inclination as in Fig 1 ) is attached. However, this is approximately independent of the Y-position on the detector, but proportional to the absorbed intensity: Δr = (I ₀ -I) * w, where w is a constant with similar dependencies as k.

Preferably, the correction of the influence of the penetration depth is performed such that a difference distance between the zero point distance of the corrected position and the zero point distance of the uncorrected position is essentially calculated from a product of the intensity measured from the respective pixel , and an intensity correction parameter. This intensity correction parameter is preferably a constant; in particular, the intensity correction parameter is not dependent on the zero point distance, at least in the linear order of the zero point distance. In particular, the intensity correction parameter is also not dependent on the intensity.

The feature that the differential distance is calculated essentially as stated means in particular that it is possible, but not necessary, for further parameters to be included in the calculation of the differential distance. Any additional contribution that is not the product of intensity and intensity correction parameter is preferably less than one third, in particular less than one fifth, of the product of intensity and intensity correction parameter.

The X-ray radiation is preferably generated by irradiating a source point of a target with electrons. The dimensions of the test specimen are preferably calculated using a magnification factor which is calculated from the quotient of the distance b of the source point from the detector and a distance a of the test specimen from the source point. The magnification factor is preferably corrected for the influence of an electron penetration depth into the target. This is also done, for example, by calculating the differential distance accordingly, as described above. In other words, both the influence of the penetration depth in the detector and the influence of a shift in the effective source point due to beam hardening can be compensated for by the calculation given above.

The magnification factor is preferably determined by measuring a preferably thin calibration body whose dimensions are known. This is done according to steps (a) to (c) according to claim 1. Subsequently, the thickness of the (pre-) filter is changed, with which the X-ray beam is filtered before it hits the hits the calibration body and is measured repeatedly. The intensity correction parameter is then calculated from a change in the magnification of the shadow image of the calibration body on the detector as a function of the intensity of the X-ray radiation. Changing the filter strength of the filter changes the intensity measured at a given pixel, as well as the x-ray spectrum.

$$\mathrm{V2}=\left(\mathrm{b}+\mathrm{d1}+\mathrm{\delta }+{\mathrm{\tau }}_{\mathrm{e}}\right)/\left(\mathrm{a}+\mathrm{d1}+{\mathrm{\tau }}_{\mathrm{e}}\right)$$

$$\mathrm{V3}=\left(\mathrm{b}+\mathrm{d2}+\mathrm{\delta }+{\mathrm{\tau }}_{\mathrm{e}}\right)/\left(\mathrm{a}+\mathrm{d2}+{\mathrm{\tau }}_{\mathrm{e}}\right)$$

$$\mathrm{V4}=\left(\mathrm{b}+\mathrm{d3}+\mathrm{\delta }+{\mathrm{\tau }}_{\mathrm{e}}\right)/\left(\mathrm{a}+\mathrm{d3}+{\mathrm{\tau }}_{\mathrm{e}}\right)$$

The exact source location and detector location shifts can be determined as follows, from which c and v can be calculated: at least four measurements must be carried out at different measured geometric magnifications V1 to V4 with different positions of the calibration body under otherwise identical conditions, so that the source location ( τ _e ) and detector location shifts (δ) can separate. A system of equations results: $$\mathrm{V1}=\left(\mathrm{b}+\mathrm{\delta }+{\mathrm{\tau }}_{\mathrm{e}}\right)/\left(\mathrm{a}+{\mathrm{\tau }}_{\mathrm{e}}\right)$$

$$\mathrm{v2}=\left(\mathrm{b}+\mathrm{d1}+\mathrm{\delta }+{\mathrm{\tau }}_{\mathrm{e}}\right)/\left(\mathrm{a}+\mathrm{d1}+{\mathrm{\tau }}_{\mathrm{e}}\right)$$

$$\mathrm{V3}=\left(\mathrm{b}+\mathrm{d2}+\mathrm{\delta }+{\mathrm{\tau }}_{\mathrm{e}}\right)/\left(\mathrm{a}+\mathrm{d2}+{\mathrm{\tau }}_{\mathrm{e}}\right)$$

$$\mathrm{V4}=\left(\mathrm{b}+\mathrm{d3}+\mathrm{\delta }+{\mathrm{\tau }}_{\mathrm{e}}\right)/\left(\mathrm{a}+\mathrm{d3}+{\mathrm{\tau }}_{\mathrm{e}}\right)$$

Measurements in more different positions are favorable, since statistically more accurate results can be calculated by overdetermining the system of equations. d1 to d3 represent displacements of the test specimen from the initial position, which can preferably be determined using a laser interferometer.

The experimentally determined magnifications V1 to V4 result from the quotient of the size of the image on the detector 14 and the known dimensions of the calibration body. The size of the image on the detector is the dimensions of the image in pixels multiplied by the known distance between two pixels. In order to take into account the offset of the calibration body from the axis of rotation, the measurements can be repeated in the 0° and 180° position of the axis and an average value can be calculated. A uniform lattice structure can preferably be used as the calibration body, the lattice constant of which is used as a material measure.

Thus, a, b, δ and τ _e are unknown variables that can be determined from the system of four equations. With the now known parameters the parameter c required for the image correction can be calculated for a given magnification. Typical values of c for the preferred parameters of the tomograph already described above are 100 μm to 300 μm, which means that the magnification changes by a relative 1×10 ^-4 to 3×10 ^-4 at a distance of one meter from detector to source, for example . By using different, preferably thin, calibration bodies made of different materials and under different conditions of the beam source, relative changes in c can be determined by comparing the geometric magnification factors with otherwise identical geometric conditions. This saves measurements in different positions to determine c for other operating conditions and test piece materials.

It is favorable if, when measuring a test object that is not a calibration body, a filter is used that filters out a maximum of 75% of the total intensity of the raw radiation. This results in a particularly short measuring time. At the same time, this requires a broad X-ray spectrum, which leads to particularly large deviations without the correction of the radiographic images according to the invention. Conversely, this means that using the method according to the invention, one can work with less filtering and thus a shorter measurement time, which represents an economic advantage.

Figur 1

schematisch einen erfindungsgemäßen Röntgen-Computertomographen zum Durchführen eines erfindungsgemäßen Verfahrens,

Figur 2

mit den Teilfiguren 2a und 2b zeigt schematisch, wie die pixelabhängigen Intensitätsdaten um den Einfluss der Eindringtiefe auf den Faktor korrigiert werden und

Figur 3

mit den Teilfiguren 3a und 3b zeigt experimentelle Daten, in denen die Abhängigkeit der Vergrößerung von der absorbierten Intensität gezeigt wird.

The invention is explained in more detail below with reference to the attached drawings. while showing

figure 1

schematically an X-ray computer tomograph according to the invention for carrying out a method according to the invention,

figure 2

with sub-figures 2a and 2b shows schematically how the pixel-dependent intensity data are corrected for the influence of the penetration depth on the factor and

figure 3

3a and 3b shows experimental data showing the dependence of magnification on absorbed intensity.

figure 1 shows a schematic view of an X-ray computer tomograph 10 according to the invention, which has an X-ray source 12 and a detector 14 . The X-ray source 12 has an electron beam source 16 for generating an electron beam 18 which is directed onto a target 20 . The target 20 consists of tungsten, for example. The electrons of the electron beam 18 have an energy of, for example, 225 kiloelectron volts. The electron beam 18 strikes the target 20 at a source point Q. An angle of incidence between the electron beam 18 and a surface of the target 20 is preferably between 15 and 30°. Alternatively, a thin target can also be irradiated from the back.

When the electrons of the electron beam 18 hit the target 20, X-rays are produced which, for the sake of simplicity, can be viewed as being composed of a plurality of X-rays 22.i. Two beam paths with index i=1, 2 are shown in the drawing. An optional filter 24 is arranged behind the target 20 in the direction of the beam, which causes a beam hardening of the X-rays 22.i. The filter consists, for example, of aluminum or copper and has a thickness d.

A test object 26 is arranged in the radiation path behind the filter 24 . The test specimen 26 includes a structure 28 to be measured, for example a bore, and the structure 28 surrounding material 30. This imaginary subdivision of the test specimen 26 into structure 28 and material 30 is only used to explain the invention and is not intended to be a restrictive statement about the type of test specimen contain.

The x-ray computed tomograph 10 preferably includes a sample holder for receiving the test object 26. The sample holder is preferably designed as a movement device 32, in particular as a rotating device for rotating the test object 26 about an axis of rotation D. The axis of rotation D has a first distance a from the source point Q.

The detector 14 is arranged behind the test object 26 in the beam direction and is at a second distance b from the source point Q. In the present case, the detector 14 has a scintillation element 34 which has a large number of microcolumns 36 . The micro columns extend perpendicularly to a detector plane E and consist, for example, of cesium iodide crystallite needles. If an X-ray quantum hits the detector 14, a flash of light is produced, which propagates along the adjacent microcolumns and thus hits a small number of photo elements 38i. It should be pointed out that the running index i is used for several objects without implying an assignment.

figure 1 shows two scenarios for a silhouette S, S' of the specimen 26 on the detector 14. For the first scenario, two X-ray beams 22.1', 22.2' are drawn in dashed lines starting from the source point Q', which correspond to the case that no filter 24 is present and the test specimen 26 is made up of the structure 28 only. As a result, there is minimal beam hardening and a high intensity measured by detector 14. Due to the low beam hardening, the effective penetration depth τ' is comparatively small. Denoted in solid line are two x-ray beams passing through the same point pairs P1 and P2 of the structure 28 where a filter 24 is present and/or the structure 28 is surrounded by a significant amount of material 30 for the second scenario. As explained in the introduction to the description, the beam is hardened and thus the effective penetration depth τ into the detector 14 is greater. The difference δ=τ−τ′ is greater than zero. The distance from the source to the detector b is therefore greater by this value and the image of the points P1 and P2 is further apart in the ratio ( b + δ ) / b than in the original image. As a second effect of the additional beam hardening by the absorber 30, the mean source location Q on the target is different. Since the target runs obliquely to the axis A, the distances a and b increase at the same time by the value τ _e , as well as a lateral offset that occurs geometrically enlarged as an offset ε in the image in the direction of the target inclination. The measurement results of the detector 14 are evaluated by means of an evaluation unit 40, which has at least one processor and one digital memory for this purpose.

Figure 2a shows how the described effect can be corrected and shows schematically a section from the detector 14 with the pixels P _x,y . The pixel P _2,3 detects a very low intensity I _2,3 = I(P _2,3 ). On the other hand, the pixels detect P _3.3 and P _2.2 have an average intensity I _3.3 and I _2.2 , respectively. For the remaining pixels P it is assumed that they detect a maximum intensity I _x,y = I _max for the sake of simplicity.

The pixel P _2.3 has a zero point distance r _2.3 from a zero point N (see figure 1 ) of the detector 14 on the optical axis A (cf. figure 1 ). The optical axis A is that line which runs through the source point Q and is perpendicular to the detector plane E along which the detector 14 extends. The source point that is detected when neither a filter nor a test object is present in the setup is used as an approximation.

Figure 2b shows that the pixel-dependent intensity data are corrected for the influence of the penetration depth τ and the source location shift τ _e , that the intensity I(P _2,3 )), i.e. the intensity measured by the pixel P ₂ , ₃ , has a new Position K is assigned. This new position K is calculated by shifting the position of the original pixel P _2,3 in the direction of the connecting line L _2,3 from the optical axis A to the original position of the _pixel P _2,3 . The new distance r'2.3 is calculated as r'2.3 = r2.3 (1+k·I _2.3 + c) with the displacement parameter k, the material-dependent constant c and the intensity I. For the sake of simplicity, here c = 0 assumed.

As Figure 2b shows schematically, this corresponds to an imaginary displacement of the position of the pixel P _2.3 relative to the original pixel pattern and thus the original coordinate system. In the same way, the corresponding displacement for all pixels P _x,y is calculated. This is indicated for the pixels P _3.3 and P _2.2 .

In a subsequent step, each pixel P _x,y is assigned a corrected intensity I' _x,y . This is done by calculating for each pixel what proportion of the area the calculated shifted intensity has at the respective pixel. Thus, the pixel P _2,3 is assigned the intensity I' _2,3 , which in the present case corresponds to 0.52 times the intensity I _2,3 , since only 52% of the black area in the area of the pixel P _{2, 3} lying, like Figure 2b can be seen. This area B is outlined in dot-dash lines in FIG. 2b. The intensity I' _3,3 assigned to the _pixel P _3,3 becomes, I' _3.3 =I _2.3* 0.22+0.82 _* I ₃₃ . This calculation is performed for all pixels P _x,y of the original image of the detector 14 . The intensity data corrected in this way result in a corrected image of the detector 14 and are then used to reconstruct a three-dimensional density image of the specimen 26 .

It should be noted that it is beneficial, but not necessary, for the intensities to be calculated for the original bitmap. It is also conceivable and covered by the invention that this pixel-by-pixel intensity correction, which redistributes intensity to other neighboring pixels, is applied to a different pixel pattern, for example a hexagonal lattice.

figure 1 shows that an enlargement factor V=b/a can be calculated from the first distance a and the second distance b, which indicates how many times larger the silhouette S, S′ appears enlarged compared to the structure 28 on the detector 14 . In order to measure a dimension, for example a height H of a recess in the test specimen 26, this magnification factor V must be known.

The magnification factors V1 to V4 and the distances a and b are explained above. Magnification is an immediate measure if one knows the dimensions of the calibration grid and assumes the pixel spacing of the detector is known (e.g. 200 µm). Firstly, these are very well known and secondly, it turns out that the detector pixel size falls out of the results, since all dimensions are measured in pixels/voxels and are related to the calibration object.

Figure 3a shows the intensity I measured by the detector as the x-axis. The intensity is changed by progressively thicker pre-filters either made of copper (circles) or aluminum (squares). The specimen 26 consists only of a structure 28 in the form of aluminum foil with a plurality of recesses placed at known positions. The y-axis indicates the magnification factor V. It can be seen that the magnification factor V decreases with increasing intensity I and increases with stronger absorption. The reason for this is the influence of increasing beam hardening on the penetration depth and the apparent position of the source point, as described above.

Figure 3b shows a diagram as in Figure 3a , where the specimen 26 consists only of a structure 28 in the form of a copper foil with a plurality of recesses placed at known positions. Reference List 10 X-ray computer tomograph P pixel 12 x-ray source Q source point 14 detector right Distance 16 electron beam source S silhouette 18 electron beam k intensity correction parameters 20 Target V magnification factor 22 X-ray w constant 24 filter 26 examinee δ difference 28 structure e shift τ effective penetration depth 30 material τe effective source point displacement 32 Sample holder, rotating device 34 scintillation element 36 microcolumn 38 photocell 40 evaluation unit a first distance A optical axis b second distance i.e filter strength D axis of rotation E detector level H Height I intensity K position L distance to the center of the image

Claims (8)

1. A method for dimensional X-ray measurement, in particular by way of computed tomography, featuring the steps:

   a) irradiating a test object (26) with non-monochromatic X-ray radiation from a virtually punctiform X-ray source (12),

   b) measuring the intensity (I) of the X-ray radiation (22) in the radiation path behind the test object (26) by means of a flat panel detector (14) which has a plurality of pixels (P) to obtain pixel-dependent intensity data (I(P)), and

   c) calculating at least one dimension (H) of the test object (26) using the pixel-dependent intensity data (I(P)),
   characterised in that

   d) the pixel-dependent intensity data (I(P)) is corrected by the influence

   - of an effective penetration depth (τ) in the flat panel detector (14) caused by beam hardening and/or

   - a displacement of the effective source location (Q) on a target (20) of the X-ray source (12) caused by beam hardening is corrected
   and
   the correction comprises the following steps:
   for at least a majority of the pixels (P)

   (i) identifying a zero point distance (r) of the pixel (P) from an optical axis (A),

   (ii) identifying the intensity (I) of the X-ray radiation measured by the pixel,

   (iii) allocating a corrected position (K') depending on the zero point distance (r) and the intensity (I), and

   (iv) calculating corrected pixel-dependent intensity data (I') from all corrected positions (K') and the corresponding intensities (I'(P)).
2. The method according to claim 1, characterised in that the corrected position (K'), the zero point (N) and the original position (K) lie on one line.
3. The method according to claim 1 or 2, characterised by the fact that

   a differential distance (_Δr) between the zero point distance (r') of the corrected position (K') and the zero point distance (r) of the uncorrected position (K) is calculated from

   a term (c+k f(I)), which contains a product (k f(I)) of a function (f) of the intensity (I), an intensity correction parameter (k) and a constant (c).
4. The method according to one of the preceding claims, characterised in that the X-ray radiation is generated by irradiating a source point (Q) of a target (20) with electrons,

   the calculation of the dimensions (H) of the test object (26) is executed using the enlargement factor (α), which depends on

   a distance (b) of the source point (Q) from the flat panel detector (14) and

   a distance (a) of the source point (Q) from the test object (26), and that

   the enlargement factor (α) is corrected by the influence of an X-ray emission spectrum that changes due to a changing electron penetration depth into the target (20).
5. The method according to claim 3, characterised by the steps:

   - measuring a test object (26) in the form of a calibration body,

   - changing a filter strength (d) of a filter (24) with which the X-ray radiation is filtered before it strikes the test object (26), and

   - calculating the intensity correction parameter (k) from a displacement of a shadow image (S, S') of the test object (26) on the flat panel detector (14), depending on the intensity (I) of the X-ray radiation.
6. The method according to claim 5, characterised in that

   - the X-ray radiation is generated by irradiating a source point (Q) of a target (20) with electrons such that raw radiation occurs, and

   - a filtering of the raw radiation by means of the filter (24) takes place, said filter filtering out a maximum of 75 % of the total intensity of the raw radiation.
7. An X-ray computed tomography scanner (10) with

   an X-ray source (12) for generating X-ray radiation,

   a flat panel detector (14), which features a plurality of pixels (P), for measuring pixel-dependent intensity data (I(P)) of the X-ray radiation,

   a movement device (32), especially a rotation device, for moving a test object (26) relative to the X-ray source (12) and the flat panel detector (14), and

   an evaluation unit (40) for calculating a three-dimensional image of the test object (26) using the pixel-dependent intensity data (I(P)),

   characterised in that

   the evaluation unit (40) is designed to automatically execute a method according to one of the above claims.
8. The X-ray computed tomography scanner according to claim 7, characterised by the fact that the evaluation unit (40) is designed to:

   obtain the corrected pixel-dependent intensity data (I'(P)) and

   calculate the three-dimensional image using the corrected pixel-independent intensity data (I'(P)).

Images